best package for matching

[Pakeezah Saadat]

Hi stata-users,

I have been trying to understand psmatch2 but it hasn't been very helpful. I don't quite understand how it works and the many variable it produces. I have also tried using teffects but because the observations are in millions and I don't have access to a multi-core stata software, so it doesn't process the command (ran it for over 12 hours - yes!).

What are my next best options in terms of resources for psmatch2 or other packages for matching (nearest neighbor matching with caliper)?

Many thanks.

[Phil Bromiley]

You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

I don't know of more efficient estimators (in a calculation speed sense), and it is best to use the estimator that really fits your data. So, I'd start by taking a small sample of the data (say 1000 observations) and debug everything using that sample. With millions of observations, you should get almost the same results with a much smaller random sample. So do everything on a random sample of say 200000 to make sure everything works and to make any model modifications you need. Then I'd set it up to run over the weekend on the full data set or I'd try to get someone with a faster, multi-core machine to run the analysis at the end.

[Pakeezah Saadat]

Hi Phil,
I didn't provide any dataset because my question was of a different nature and perhaps not suitable for this forum which is more for debugging codes. I did take your advice on trying the codes on a smaller random dataset and turns out coarsened exact matching works quite better.

I have a multi-core machine but I am using Stata 14.2 which is not MP. So, need to find better stata.
Thanks for the advice.